EmoSSR: behavior

setwd(paste0(getwd(), "/behavior")) # set working subdir

behav.exps <- c(
  list.files(pattern = ".csv")[2],
  list.files(pattern = ".csv")[3],
  list.files(pattern = ".csv")[1]
)

behav.data <- behav.exps %>%
  map_df(~ read_csv(.)) %>%
  mutate(
    experiment = factor(
      c(
        rep("pilot1", 15 * 6),
        rep("pilot2", 16 * 12),
        rep("exp1", 31 * 12)
      ),
      levels = c("pilot1", "pilot2", "exp1")
    ),
    participant = as.factor(subjects),
    orientation = as.factor(
      case_when(
        grepl("up", cond) ~ "upright",
        grepl("dw", cond) ~ "inverted"
      )
    ),
    variability = as.factor(
      case_when(
        grepl("_m", cond) ~ "high",
        grepl("_f", cond) ~ "low",
        TRUE ~ "NA"
      )
    ),
    regularity = as.factor(
      case_when(
        grepl("A_", cond) | grepl("a_", cond) ~ "angry",
        grepl("N_", cond) | grepl("n_", cond) ~ "neutral",
        grepl("I_", cond) | grepl("i_", cond) ~ "irregular"
      )
    )
  ) %>%
  unite(condition, c(orientation, regularity, variability), sep = "_", remove = FALSE) %>%
  dplyr::select(experiment, participant, orientation, regularity, variability, condition, RTs = "rt")

# IMPUTATION MISSING DATA
length(which(is.na(behav.data$RTs))) # number of missing values to impute

# imputation (in different data frame)
temp.behav.data <- mice(behav.data,
  m = 1, # number of multiple imputations
  maxit = 1000, # maximum number of iterations
  meth = "pmm", # imputation method: Predictive mean matching (see ?mice)
  seed = seed.smorfia # RNG seed for reproducibility
)

# graphical comparison of observed (in blue) and imputed data (in pink)
# if the imputation was successful, the two distributions should overlap
densityplot(temp.behav.data)
# despite the low number of data points, the imputation was reasonably close to the observed data

# insert imputed values in original data frame
behav.data <- complete(temp.behav.data)

# verify that imputation occurred
length(which(is.na(behav.data$RTs)))

write_csv(behav.data, # save imputed data
  path = "RTs_imputed.csv"
)

rm(behav.exps, temp.behav.data) # remove unused variables

Behavioral data were collected in 2 pilots and 1 main experiment, whereas in the last experiment participants were asked to passively view the stream of faces.

# setwd(paste0(getwd(), "/behavior")) # set working subdir

behav.data <- read_csv("RTs_imputed.csv", # read imputed data
  col_types = cols(
    experiment = "f",
    participant = "f",
    orientation = "f",
    regularity = "f",
    variability = "f",
    condition = "f"
  )
) %>%
  mutate(
    experiment = factor(experiment, levels = c("pilot1", "pilot2", "exp1"))
  )

summary.behav.data <- summarySEwithin(
  data = behav.data,
  measurevar = "RTs",
  betweenvars = "experiment",
  withinvars = c("orientation", "regularity", "variability"),
  idvar = "participant",
  na.rm = TRUE
) %>%
  mutate(
    CI.95.low = RTs - ci,
    CI.95.high = RTs + ci
  ) %>%
  dplyr::select(-c(sd, se, ci))

kable(summary.behav.data, digits = 2)

experiment	orientation	regularity	variability	N	RTs	CI.95.low	CI.95.high
pilot1	upright	angry	NA	15	519.4	514.9	523.8
pilot1	upright	neutral	NA	15	520.5	513.7	527.2
pilot1	upright	irregular	NA	15	513.9	505.8	522.0
pilot1	inverted	angry	NA	15	518.6	510.6	526.7
pilot1	inverted	neutral	NA	15	519.7	511.9	527.5
pilot1	inverted	irregular	NA	15	524.8	516.8	532.8
pilot2	upright	angry	low	16	513.9	496.5	531.3
pilot2	upright	angry	high	16	520.0	487.0	552.9
pilot2	upright	neutral	low	16	489.1	460.5	517.7
pilot2	upright	neutral	high	16	531.7	501.9	561.5
pilot2	upright	irregular	low	16	488.2	451.2	525.2
pilot2	upright	irregular	high	16	545.0	476.4	613.6
pilot2	inverted	angry	low	16	526.1	501.5	550.8
pilot2	inverted	angry	high	16	507.1	476.3	537.9
pilot2	inverted	neutral	low	16	491.3	458.3	524.4
pilot2	inverted	neutral	high	16	511.0	476.4	545.5
pilot2	inverted	irregular	low	16	560.0	511.9	608.1
pilot2	inverted	irregular	high	16	550.4	496.1	604.7
exp1	upright	angry	low	31	516.0	476.5	555.4
exp1	upright	angry	high	31	502.2	461.3	543.2
exp1	upright	neutral	low	31	509.4	465.4	553.5
exp1	upright	neutral	high	31	509.0	477.2	540.8
exp1	upright	irregular	low	31	481.0	447.2	514.8
exp1	upright	irregular	high	31	560.0	516.6	603.5
exp1	inverted	angry	low	31	486.0	443.3	528.8
exp1	inverted	angry	high	31	523.0	477.1	569.0
exp1	inverted	neutral	low	31	549.0	509.5	588.6
exp1	inverted	neutral	high	31	512.1	474.9	549.2
exp1	inverted	irregular	low	31	536.6	492.3	581.0
exp1	inverted	irregular	high	31	549.3	504.7	593.9

behav.data <- behav.data %>%
  unite("cond4plot", c(orientation, regularity), sep = " ", remove = FALSE) %>%
  mutate(cond4plot = factor(cond4plot,
    levels = c(
      "upright angry", "upright neutral", "upright irregular",
      "inverted angry", "inverted neutral", "inverted irregular"
    )
  ))

# pilot1 does not have the column "variability", which messes up all the other plots...
# that's why they are created separately instead of calling facet_wrap(~ experiment, scales = "free")

# pilot1
behav.plot.pilot1 <-
  behav.data %>%
  filter(experiment == "pilot1") %>%
  ggplot(aes(
    x = cond4plot,
    y = RTs,
    color = cond4plot,
    fill = cond4plot
  )) +
  geom_pirate(
    bars = FALSE,
    cis = TRUE,
    lines = TRUE, lines_params = list(color = "black"),
    points = TRUE, points_params = list(shape = 16, color = "black", size = 5, alpha = .2),
    violins = TRUE, violins_params = list(size = 1),
    show.legend = FALSE
  ) +
  scale_fill_viridis_d() +
  scale_color_viridis_d() +
  scale_y_continuous(
    limits = c(0, 800),
    breaks = seq(0, 800, 100)
  ) +
  coord_cartesian(ylim = c(200, 800)) +
  geom_hline(
    yintercept = seq(0, 800, 100),
    linetype = "dotted",
    colour = "#999999",
    size = .8,
    alpha = .5
  ) +
  labs(
    x = "",
    y = "ms"
  ) +
  ggtitle("pilot 1") +
  theme_EmoSSR +
  theme(legend.position = "right")

# pilot2
behav.plot.pilot2 <-
  behav.data %>%
  filter(experiment == "pilot2") %>%
  ggplot(aes(
    x = cond4plot,
    y = RTs,
    color = variability,
    fill = variability
  )) +
  geom_pirate(
    bars = FALSE,
    cis = TRUE,
    lines = TRUE, lines_params = list(color = "black"),
    points = TRUE, points_params = list(shape = 16, size = 5, alpha = .2),
    violins = TRUE, violins_params = list(size = 1),
    show.legend = TRUE
  ) +
  scale_fill_viridis_d(option = "cividis") +
  scale_color_viridis_d(option = "cividis") +
  scale_y_continuous(
    limits = c(0, 800),
    breaks = seq(0, 800, 100)
  ) +
  coord_cartesian(ylim = c(200, 800)) +
  geom_hline(
    yintercept = seq(0, 800, 100),
    linetype = "dotted",
    colour = "#999999",
    size = .8,
    alpha = .5
  ) +
  labs(
    x = "",
    y = "ms"
  ) +
  ggtitle("pilot 2") +
  theme_EmoSSR +
  theme(legend.position = "right")

# exp1
behav.plot.exp1 <-
  behav.data %>%
  filter(experiment == "exp1") %>%
  ggplot(aes(
    x = cond4plot,
    y = RTs,
    color = variability,
    fill = variability
  )) +
  geom_pirate(
    bars = FALSE,
    cis = TRUE,
    lines = TRUE, lines_params = list(color = "black"),
    points = TRUE, points_params = list(shape = 16, size = 5, alpha = .2),
    violins = TRUE, violins_params = list(size = 1),
    show.legend = TRUE
  ) +
  scale_fill_viridis_d(option = "cividis") +
  scale_color_viridis_d(option = "cividis") +
  scale_y_continuous(
    limits = c(0, 800),
    breaks = seq(0, 800, 100)
  ) +
  coord_cartesian(ylim = c(200, 800)) +
  geom_hline(
    yintercept = seq(0, 800, 100),
    linetype = "dotted",
    colour = "#999999",
    size = .8,
    alpha = .5
  ) +
  labs(
    x = "",
    y = "ms"
  ) +
  ggtitle("experiment 1") +
  theme_EmoSSR +
  theme(legend.position = "right")

plot_grid(behav.plot.pilot1, behav.plot.pilot2, behav.plot.exp1,
          ncol = 1)

# ### save plots to file
# # pilot1
# file.behav.plot.pilot1 <- behav.plot.pilot1 +
#   scale_y_continuous(
#     limits = c(0, 800),
#     breaks = seq(0, 800, 50)
#   ) +
#   geom_hline(
#     yintercept = seq(0, 800, 50),
#     linetype = "dotted",
#     colour = "#999999",
#     size = .8,
#     alpha = .5
#   ) +
#   coord_cartesian(ylim = c(300, 600))
# 
# save_plot(paste0(getwd(), "/pilot1_behavior.svg"),
#           file.behav.plot.pilot1,
#           base_aspect_ratio = 1,
#           base_width = 11
#           )
# 
# # pilot2
# file.behav.plot.pilot2 <- behav.plot.pilot2 +
#   coord_cartesian(ylim = c(300, 800))
# 
# save_plot(paste0(getwd(), "/pilot2_behavior.svg"),
#           file.behav.plot.pilot2,
#           base_aspect_ratio = 1,
#           base_width = 11
#           )
# 
# # exp1
# save_plot(paste0(getwd(), "/exp1_behavior.svg"),
#           behav.plot.exp1,
#           base_aspect_ratio = 1,
#           base_width = 12
#           )

For the statistical analyses, we fit Bayesian multilevel models using brms. The constant effect was condition. Intercepts and slopes were allowed to vary as a function of participant.

Pilot 1

• flickering frequency: 15 Hz
• regularity: 5 Hz
• 9 face identities
• task: detect dots on faces

# subset data
behav.pilot1 <- behav.data %>%
  filter(experiment == "pilot1") %>%
  dplyr::select(-variability) %>%
  droplevels() %>%
  mutate(condition = as.factor(gsub("_NA", "", condition)))

pilot1.model <- brm(RTs ~ 0 + condition # constant effects
+ (0 + condition | participant), # varying effects
data = behav.pilot1, # subset data current experiment
family = gaussian(), # likelihood function
prior = weak.priors, # priors
sample_prior = TRUE, # draw samples from the prior distribution
inits = "0", # initial parameter values in the chains
control = list(
  adapt_delta = .99, # parameters of NUTS sampler
  max_treedepth = 15
),
chains = num.chains, # number of chains
iter = num.iter, # number of iterations
warmup = num.warmup, # number of warm-up samples
thin = num.thin, # thinning rate
algorithm = "sampling", # type of sampling algorithm
cores = num.chains, # number of processor cores to use when running chains in parallel
seed = seed.smorfia # RNG seed
)

saveRDS(pilot1.model, file = "pilot1_model.rds", compress = "gzip") # save results

# subset data
behav.pilot1 <- behav.data %>%
  filter(experiment == "pilot1") %>%
  dplyr::select(-variability) %>%
  droplevels() %>%
  mutate(condition = as.factor(gsub("_NA", "", condition)))

# load model
pilot1.model <- readRDS("pilot1_model.rds")

# summary posterior distributions & diagnostics
summary.pilot1.model <-
  describe_posterior(pilot1.model,
    centrality = "mean",
    ci = 0.95,
    ci_method = "hdi",
    test = NULL,
    diagnostic = "all"
  ) %>%
  mutate(
    Parameter = gsub("b_condition", "", Parameter),
    Parameter2 = gsub("_", " ", Parameter),
    condition = factor(Parameter2,
      levels =
        c(
          "upright angry", "upright neutral", "upright irregular",
          "inverted angry", "inverted neutral", "inverted irregular"
        )
    )
  ) %>%
  dplyr::select(Parameter, Mean, HDI95_Low = CI_low, HDI95_High = CI_high, Estimated_Sample_Size = ESS, MonteCarlo_StdErr = MCSE)

kable(summary.pilot1.model, digits = 2)

Parameter	Mean	HDI95_Low	HDI95_High	Estimated_Sample_Size	MonteCarlo_StdErr
inverted_angry	437.1	420.4	452.2	7273	0.09
inverted_irregular	443.3	427.5	459.4	8041	0.09
inverted_neutral	438.1	422.9	453.1	8575	0.08
upright_angry	437.9	426.4	449.4	8374	0.06
upright_irregular	432.3	418.9	445.4	10495	0.07
upright_neutral	439.0	424.1	453.0	8191	0.08

# MCMC chain convergence
pilot1.model %>%
  spread_draws(`b_.*`, regex = TRUE) %>%
  dplyr::select(
    "Chain" = ".chain",
    "upright angry" = "b_conditionupright_angry",
    "upright neutral" = "b_conditionupright_neutral",
    "upright irregular" = "b_conditionupright_irregular",
    "inverted angry" = "b_conditioninverted_angry",
    "inverted neutral" = "b_conditioninverted_neutral",
    "inverted irregular" = "b_conditioninverted_irregular"
  ) %>%
  as.data.frame() %>%
  mcmc_trace(facet_args = list(nrow = 3)) +
  ggtitle("chain convergence") +
  theme_EmoSSR +
  theme(legend.position = "none")

# posterior predictive checks
pilot1.model %>%
  posterior_predict(draws = 500) %>%
  ppc_stat_grouped(
    y = pull(behav.pilot1, RTs),
    group = pull(behav.pilot1, cond4plot),
    stat = "mean"
  ) +
  ggtitle("posterior predictive checks") +
  theme_EmoSSR +
  theme(legend.position = "none")

pilot1.posterior.test <- pilot1.model %>%
  spread_draws(`b_.*`, regex = TRUE) %>%
  dplyr::select(
    "upright angry" = "b_conditionupright_angry",
    "upright neutral" = "b_conditionupright_neutral",
    "upright irregular" = "b_conditionupright_irregular",
    "inverted angry" = "b_conditioninverted_angry",
    "inverted neutral" = "b_conditioninverted_neutral",
    "inverted irregular" = "b_conditioninverted_irregular"
  ) %>%
  mutate(
    ### orientation (upright minus inverted)
    # angry
    diff.angry.uprightVSinverted = `upright angry` - `inverted angry`,
    # neutral
    diff.neutral.uprightVSinverted = `upright neutral` - `inverted neutral`,
    # irregular
    diff.irregular.uprightVSinverted = `upright irregular` - `inverted irregular`,
    ### regularity
    ## upright
    # angry minus neutral
    diff.upright.angryVSneutral = `upright angry` - `upright neutral`,
    # angry minus irregular
    diff.upright.angryVSirregular = `upright angry` - `upright irregular`,
    # neutral minus irregular
    diff.upright.neutralVSirregular = `upright neutral` - `upright irregular`,
    ## inverted
    # angry minus neutral
    diff.inverted.angryVSneutral = `inverted angry` - `inverted neutral`,
    # angry minus irregular
    diff.inverted.angryVSirregular = `inverted angry` - `inverted irregular`,
    # neutral minus irregular
    diff.inverted.neutralVSirregular = `inverted neutral` - `inverted irregular`
  ) %>%
  gather(key = "condition", value = "RTs") %>%
  filter(condition == c("diff.angry.uprightVSinverted",
                        "diff.neutral.uprightVSinverted",
                        "diff.irregular.uprightVSinverted",
                        "diff.upright.angryVSneutral",
                        "diff.upright.angryVSirregular",
                        "diff.upright.neutralVSirregular",
                        "diff.inverted.angryVSneutral",
                        "diff.inverted.angryVSirregular",
                        "diff.inverted.neutralVSirregular")) %>%
  droplevels() %>%
  mutate(condition = factor(condition,
                            levels = c("diff.angry.uprightVSinverted",
                                       "diff.neutral.uprightVSinverted",
                                       "diff.irregular.uprightVSinverted",
                                       "diff.upright.angryVSneutral",
                                       "diff.upright.angryVSirregular",
                                       "diff.upright.neutralVSirregular",
                                       "diff.inverted.angryVSneutral",
                                       "diff.inverted.angryVSirregular",
                                       "diff.inverted.neutralVSirregular"))
  )

# summary
summary.pilot1.posterior.test <- pilot1.posterior.test %>%
  group_by(condition) %>%
  dplyr::summarize(
    mean = mean(RTs),
    hdi.95.low = hdi(RTs)[1],
    hdi.95.high = hdi(RTs)[2],
    evidence.ratio = ifelse(mean < 0, 
                            length(which(RTs < 0)) / length(which(RTs > 0)),
                            length(which(RTs > 0)) / length(which(RTs < 0))
    )
  ) %>%
  ungroup()

kable(summary.pilot1.posterior.test, digits = 2)

condition	mean	hdi.95.low	hdi.95.high	evidence.ratio
diff.angry.uprightVSinverted	0.48	-13.05	15.83	1.11
diff.neutral.uprightVSinverted	0.79	-16.08	17.58	1.17
diff.irregular.uprightVSinverted	-11.13	-25.50	6.98	11.61
diff.upright.angryVSneutral	-1.11	-12.50	10.74	1.38
diff.upright.angryVSirregular	5.59	-5.15	17.56	5.49
diff.upright.neutralVSirregular	6.76	-7.88	21.54	4.97
diff.inverted.angryVSneutral	-0.59	-16.85	16.31	1.11
diff.inverted.angryVSirregular	-6.34	-23.32	11.93	3.36
diff.inverted.neutralVSirregular	-5.23	-17.23	6.12	4.61

### Paired comparisons ###

par(mfrow = c(3, 3))

for(i in 1:length(levels(pilot1.posterior.test$condition))) {
  
  temp.data <- filter(pilot1.posterior.test, condition == levels(condition)[i])
  
  plotPost(
    pull(temp.data, RTs), 
    credMass = 0.95,
    compVal = 0,
    ROPE = NULL,
    xlab = "", 
    col = "#b3cde0", 
    showCurve = FALSE,
    cex = 1,
    main = levels(temp.data$condition)[i]
  )
  
}

Take-home message:

• irregular: slightly slower RTs for inverted compared to upright faces

---

---

Pilot 2

• flickering frequency: 15 Hz
• regularity: 5 Hz
• 2 face identities
• task: detect dots on faces

# subset data
behav.pilot2 <- behav.data %>%
  filter(experiment == "pilot2")

pilot2.model <- brm(RTs ~ 0 + condition
+ (0 + condition | participant),
data = behav.pilot2,
family = gaussian(),
prior = weak.priors,
sample_prior = TRUE,
inits = "random",
control = list(
  adapt_delta = .99,
  max_treedepth = 15
),
chains = num.chains,
iter = num.iter,
warmup = num.warmup,
thin = num.thin,
algorithm = "sampling",
cores = num.chains,
seed = seed.smorfia
)

saveRDS(pilot2.model, file = "pilot2_model.rds", compress = "gzip")

# subset data
behav.pilot2 <- behav.data %>%
  filter(experiment == "pilot2")

# load model
pilot2.model <- readRDS("pilot2_model.rds")

# summary posterior distributions & diagnostics
summary.pilot2.model <-
  describe_posterior(pilot2.model,
    centrality = "mean",
    ci = 0.95,
    ci_method = "hdi",
    test = NULL,
    diagnostic = "all"
  ) %>%
  mutate(
    Parameter = gsub("b_condition", "", Parameter),
    Parameter2 = gsub("_", " ", Parameter),
    condition = factor(Parameter2,
      levels =
        c(
          "upright angry high", "upright neutral high", "upright irregular high",
          "upright angry low", "upright neutral low", "upright irregular low",
          "inverted angry high", "inverted neutral high", "inverted irregular high",
          "inverted angry low", "inverted neutral low", "inverted irregular low"
        )
    )
  ) %>%
  dplyr::select(Parameter, Mean, HDI95_Low = CI_low, HDI95_High = CI_high, Estimated_Sample_Size = ESS, MonteCarlo_StdErr = MCSE)

kable(summary.pilot2.model, digits = 2)

	Parameter	Mean	HDI95_Low	HDI95_High	Estimated_Sample_Size	MonteCarlo_StdErr
10	upright_angry_low	547.2	520.6	575.0	12524	0.12
16	upright_neutral_low	522.4	492.5	552.5	12025	0.14
13	upright_irregular_low	521.6	490.4	554.1	12563	0.15
14	upright_irregular_low	521.6	490.4	554.1	12563	0.15
2	inverted_angry_low	559.5	531.6	587.3	13287	0.12
8	inverted_neutral_low	525.1	494.6	555.8	12763	0.14
5	inverted_irregular_low	592.8	549.1	637.3	13369	0.19
6	inverted_irregular_low	592.8	549.1	637.3	13369	0.19
9	upright_angry_high	553.2	525.0	580.7	12673	0.13
15	upright_neutral_high	564.9	536.2	594.1	13611	0.13
11	upright_irregular_high	577.6	510.7	642.1	12878	0.30
12	upright_irregular_high	577.6	510.7	642.1	12878	0.30
1	inverted_angry_high	540.5	510.4	570.6	13191	0.13
7	inverted_neutral_high	544.4	512.2	577.3	13380	0.14
3	inverted_irregular_high	583.2	528.8	634.3	13746	0.23
4	inverted_irregular_high	583.2	528.8	634.3	13746	0.23

# MCMC chains
pilot2.model %>%
  spread_draws(`b_.*`, regex = TRUE) %>%
  dplyr::select(
    "Chain" = ".chain",
    "upright angry high" = "b_conditionupright_angry_high",
    "upright neutral high" = "b_conditionupright_neutral_high",
    "upright irregular high" = "b_conditionupright_irregular_high",
    "upright angry low" = "b_conditionupright_angry_low",
    "upright neutral low" = "b_conditionupright_neutral_low",
    "upright irregular low" = "b_conditionupright_irregular_low",
    "inverted angry high" = "b_conditioninverted_angry_high",
    "inverted neutral high" = "b_conditioninverted_neutral_high",
    "inverted irregular high" = "b_conditioninverted_irregular_high",
    "inverted angry low" = "b_conditioninverted_angry_low",
    "inverted neutral low" = "b_conditioninverted_neutral_low",
    "inverted irregular low" = "b_conditioninverted_irregular_low"
  ) %>%
  as.data.frame() %>%
  mcmc_trace(facet_args = list(ncol = 3)) +
  geom_vline(
    xintercept = seq(0, 2000, 500),
    linetype = "dotted",
    colour = "#999999",
    size = .8,
    alpha = .5
  ) +
  ggtitle("chain convergence") +
  theme_EmoSSR

# posterior predictive checks
pilot2.model %>%
  posterior_predict(draws = 500) %>%
  ppc_stat_grouped(
    y = pull(behav.pilot2, RTs),
    group = pull(behav.pilot2, condition),
    stat = "mean"
  ) +
  ggtitle("posterior predictive checks") +
  theme_EmoSSR +
  theme(legend.position = "none")

# posterior densities
pilot2.model %>%
  gather_draws(`b_.*`, regex = TRUE) %>%
  mutate(condition = recode(factor(.variable),
    "b_conditionupright_angry_high" = "upright angry high",
    "b_conditionupright_neutral_high" = "upright neutral high",
    "b_conditionupright_irregular_high" = "upright irregular high",
    "b_conditionupright_angry_low" = "upright angry low",
    "b_conditionupright_neutral_low" = "upright neutral low",
    "b_conditionupright_irregular_low" = "upright irregular low",
    "b_conditioninverted_angry_high" = "inverted angry high",
    "b_conditioninverted_neutral_high" = "inverted neutral high",
    "b_conditioninverted_irregular_high" = "inverted irregular high",
    "b_conditioninverted_angry_low" = "inverted angry low",
    "b_conditioninverted_neutral_low" = "inverted neutral low",
     "b_conditioninverted_irregular_low" = "inverted irregular low"
  )) %>%
  dplyr::select(condition, RTs = .value) %>%
  mutate(condition = factor(condition,
    levels = c(
          "upright angry high", "upright neutral high", "upright irregular high",
          "upright angry low", "upright neutral low", "upright irregular low",
          "inverted angry high", "inverted neutral high", "inverted irregular high",
          "inverted angry low", "inverted neutral low", "inverted irregular low"
    )
  )) %>%
  ggplot(aes(y = condition, x = RTs)) +
  geom_halfeyeh(fill = "#b3cde0", .width = .95) +
  ggtitle("posterior densities") +
  theme_minimal(base_size = 18) +
  theme(
    strip.text = element_text(
      hjust = .5,
      size = 24
    ),
    plot.title = element_text(size = 22, hjust = .5)
  )

### observed vs. predicted data ###
behav.pilot2 %>%
  mutate(predicted = predict(pilot2.model)[, 1]) %>%
  dplyr::rename(observed = RTs) %>%
  gather(key = data_type, value = RTs, observed:predicted) %>%
  ggplot(aes(
    x = cond4plot,
    y = RTs,
    color = variability,
    fill = variability
  )) +
  geom_pirate(
    bars = FALSE,
    cis = TRUE,
    lines = TRUE, lines_params = list(color = "black"),
    points = TRUE, points_params = list(shape = 16, size = 5, alpha = .2),
    violins = TRUE, violins_params = list(size = 1),
    show.legend = TRUE
  ) +
  scale_fill_viridis_d(option = "cividis") +
  scale_color_viridis_d(option = "cividis") +
  scale_y_continuous(
    limits = c(0, 800),
    breaks = seq(0, 800, 100)
  ) +
  coord_cartesian(ylim = c(200, 800)) +
  geom_hline(
    yintercept = seq(0, 800, 100),
    linetype = "dotted",
    colour = "#999999",
    size = .8,
    alpha = .5
  ) +
  labs(
    x = "",
    y = "ms"
  ) +
  facet_wrap(~ data_type, scales = "free") +
  ggtitle("pilot 2") +
  theme_EmoSSR

pilot2.posterior.test <- pilot2.model %>%
  spread_draws(`b_.*`, regex = TRUE) %>%
  dplyr::select(
    "upright angry high" = "b_conditionupright_angry_high",
    "upright neutral high" = "b_conditionupright_neutral_high",
    "upright irregular high" = "b_conditionupright_irregular_high",
    "upright angry low" = "b_conditionupright_angry_low",
    "upright neutral low" = "b_conditionupright_neutral_low",
    "upright irregular low" = "b_conditionupright_irregular_low",
    "inverted angry high" = "b_conditioninverted_angry_high",
    "inverted neutral high" = "b_conditioninverted_neutral_high",
    "inverted irregular high" = "b_conditioninverted_irregular_high",
    "inverted angry low" = "b_conditioninverted_angry_low",
    "inverted neutral low" = "b_conditioninverted_neutral_low",
    "inverted irregular low" = "b_conditioninverted_irregular_low"
  ) %>%
  mutate(
    ### orientation
    # angry high
    diff.high.angry.uprightVSinverted = `upright angry high` - `inverted angry high`,
    # neutral high
    diff.high.neutral.uprightVSinverted = `upright neutral high` - `inverted neutral high`,
    # irregular high
    diff.high.irregular.uprightVSinverted = `upright irregular high` - `inverted irregular high`,
    # angry low
    diff.low.angry.uprightVSinverted = `upright angry low` - `inverted angry low`,
    # neutral high
    diff.low.neutral.uprightVSinverted = `upright neutral low` - `inverted neutral low`,
    # irregular high
    diff.low.irregular.uprightVSinverted = `upright irregular low` - `inverted irregular low`,
    ### regularity
    ## upright high
    # angry minus neutral
    diff.high.upright.angryVSneutral = `upright angry high` - `upright neutral high`,
    # angry minus irregular
    diff.high.upright.angryVSirregular = `upright angry high` - `upright irregular high`,
    # neutral minus irregular
    diff.high.upright.neutralVSirregular = `upright neutral high` - `upright irregular high`,
    ## upright low
    # angry minus neutral
    diff.low.upright.angryVSneutral = `upright angry low` - `upright neutral low`,
    # angry minus irregular
    diff.low.upright.angryVSirregular = `upright angry low` - `upright irregular low`,
    # neutral minus irregular
    diff.low.upright.neutralVSirregular = `upright neutral low` - `upright irregular low`,
    ## inverted high
    # angry minus neutral
    diff.high.inverted.angryVSneutral = `inverted angry high` - `inverted neutral high`,
    # angry minus irregular
    diff.high.inverted.angryVSirregular = `inverted angry high` - `inverted irregular high`,
    # neutral minus irregular
    diff.high.inverted.neutralVSirregular = `inverted neutral high` - `inverted irregular high`,
    ## inverted low
    # angry minus neutral
    diff.low.inverted.angryVSneutral = `inverted angry low` - `inverted neutral low`,
    # angry minus irregular
    diff.low.inverted.angryVSirregular = `inverted angry low` - `inverted irregular low`,
    # neutral minus irregular
    diff.low.inverted.neutralVSirregular = `inverted neutral low` - `inverted irregular low`,
    ### variability
    ## upright angry
    diff.upright.angry.highVSlow = `upright angry high` - `upright angry low`,
    ## upright neutral
    diff.upright.neutral.highVSlow = `upright neutral high` - `upright neutral low`,
    ## upright irregular
    diff.upright.irregular.highVSlow = `upright irregular high` - `upright irregular low`,
    ## inverted angry
    diff.inverted.angry.highVSlow = `inverted angry high` - `inverted angry low`,
    ## inverted neutral
    diff.inverted.neutral.highVSlow = `inverted neutral high` - `inverted neutral low`,
    ## inverted irregular
    diff.inverted.irregular.highVSlow = `inverted irregular high` - `inverted irregular low`
  ) %>%
  gather(key = "condition", value = "RTs") %>%
  filter(condition == c(
    "diff.high.angry.uprightVSinverted",
    "diff.high.neutral.uprightVSinverted",
    "diff.high.irregular.uprightVSinverted",
    "diff.low.angry.uprightVSinverted",
    "diff.low.neutral.uprightVSinverted",
    "diff.low.irregular.uprightVSinverted",
    "diff.high.upright.angryVSneutral",
    "diff.high.upright.angryVSirregular",
    "diff.high.upright.neutralVSirregular",
    "diff.low.upright.angryVSneutral",
    "diff.low.upright.angryVSirregular",
    "diff.low.upright.neutralVSirregular",
    "diff.high.inverted.angryVSneutral",
    "diff.high.inverted.angryVSirregular",
    "diff.high.inverted.neutralVSirregular",
    "diff.low.inverted.angryVSneutral",
    "diff.low.inverted.angryVSirregular",
    "diff.low.inverted.neutralVSirregular",
    "diff.upright.angry.highVSlow",
    "diff.upright.neutral.highVSlow",
    "diff.upright.irregular.highVSlow",
    "diff.inverted.angry.highVSlow",
    "diff.inverted.neutral.highVSlow",
    "diff.inverted.irregular.highVSlow"
  )) %>%
  droplevels() %>%
  mutate(condition = factor(condition,
    levels = c(
      "diff.high.angry.uprightVSinverted",
      "diff.high.neutral.uprightVSinverted",
      "diff.high.irregular.uprightVSinverted",
      "diff.low.angry.uprightVSinverted",
      "diff.low.neutral.uprightVSinverted",
      "diff.low.irregular.uprightVSinverted",
      "diff.high.upright.angryVSneutral",
      "diff.high.upright.angryVSirregular",
      "diff.high.upright.neutralVSirregular",
      "diff.low.upright.angryVSneutral",
      "diff.low.upright.angryVSirregular",
      "diff.low.upright.neutralVSirregular",
      "diff.high.inverted.angryVSneutral",
      "diff.high.inverted.angryVSirregular",
      "diff.high.inverted.neutralVSirregular",
      "diff.low.inverted.angryVSneutral",
      "diff.low.inverted.angryVSirregular",
      "diff.low.inverted.neutralVSirregular",
      "diff.upright.angry.highVSlow",
      "diff.upright.neutral.highVSlow",
      "diff.upright.irregular.highVSlow",
      "diff.inverted.angry.highVSlow",
      "diff.inverted.neutral.highVSlow",
      "diff.inverted.irregular.highVSlow"
    )
  ))

# summary
summary.pilot2.posterior.test <- pilot2.posterior.test %>%
  group_by(condition) %>%
  dplyr::summarize(
    mean = mean(RTs),
    hdi.95.low = hdi(RTs)[1],
    hdi.95.high = hdi(RTs)[2],
    evidence.ratio = ifelse(mean < 0, 
                            length(which(RTs < 0)) / length(which(RTs > 0)),
                            length(which(RTs > 0)) / length(which(RTs < 0))
    )
  ) %>%
  ungroup()

kable(summary.pilot2.posterior.test, digits = 2)

condition	mean	hdi.95.low	hdi.95.high	evidence.ratio
diff.high.angry.uprightVSinverted	14.26	-26.02	58.01	2.81
diff.high.neutral.uprightVSinverted	20.78	-21.85	59.68	5.54
diff.high.irregular.uprightVSinverted	-3.52	-71.89	93.20	1.18
diff.low.angry.uprightVSinverted	-13.03	-47.47	26.28	3.17
diff.low.neutral.uprightVSinverted	-3.75	-45.00	34.61	1.33
diff.low.irregular.uprightVSinverted	-71.78	-123.64	-14.71	132.20
diff.high.upright.angryVSneutral	-11.59	-53.36	29.25	2.75
diff.high.upright.angryVSirregular	-24.49	-88.59	56.67	3.02
diff.high.upright.neutralVSirregular	-15.05	-93.75	48.13	2.12
diff.low.upright.angryVSneutral	25.18	-12.06	67.21	9.42
diff.low.upright.angryVSirregular	25.67	-16.16	62.58	7.54
diff.low.upright.neutralVSirregular	0.97	-40.96	46.95	1.11
diff.high.inverted.angryVSneutral	-3.69	-44.22	35.52	1.25
diff.high.inverted.angryVSirregular	-41.42	-104.62	21.02	9.92
diff.high.inverted.neutralVSirregular	-40.00	-109.86	17.40	8.81
diff.low.inverted.angryVSneutral	32.84	-10.58	70.07	18.06
diff.low.inverted.angryVSirregular	-32.81	-84.74	20.69	7.13
diff.low.inverted.neutralVSirregular	-68.07	-128.91	-24.05	94.29
diff.upright.angry.highVSlow	4.77	-37.43	42.06	1.59
diff.upright.neutral.highVSlow	42.07	4.62	82.04	50.31
diff.upright.irregular.highVSlow	57.46	-14.80	129.75	14.88
diff.inverted.angry.highVSlow	-19.09	-64.26	18.85	4.41
diff.inverted.neutral.highVSlow	19.72	-21.72	67.06	4.17
diff.inverted.irregular.highVSlow	-9.55	-81.61	61.61	1.63

### Paired comparisons ###

### orientation
pilot2.posterior.test.orientation <- filter(
  pilot2.posterior.test,
  condition == c(
    "diff.high.angry.uprightVSinverted",
    "diff.high.neutral.uprightVSinverted",
    "diff.high.irregular.uprightVSinverted",
    "diff.low.angry.uprightVSinverted",
    "diff.low.neutral.uprightVSinverted",
    "diff.low.irregular.uprightVSinverted"
  )
) %>%
  droplevels()

par(mfrow = c(2, 3))

for (i in 1:length(levels(pilot2.posterior.test.orientation$condition))) {
  temp.data <- filter(pilot2.posterior.test.orientation, condition == levels(condition)[i])

  plotPost(
    pull(temp.data, RTs),
    credMass = 0.95,
    compVal = 0,
    ROPE = NULL,
    xlab = "",
    col = "#b3cde0",
    showCurve = FALSE,
    cex = 1,
    main = levels(temp.data$condition)[i]
  )
}

### regularity
pilot2.posterior.test.regularity <- filter(
  pilot2.posterior.test,
  condition == c(
    "diff.high.upright.angryVSneutral",
    "diff.high.upright.angryVSirregular",
    "diff.high.upright.neutralVSirregular",
    "diff.low.upright.angryVSneutral",
    "diff.low.upright.angryVSirregular",
    "diff.low.upright.neutralVSirregular",
    "diff.high.inverted.angryVSneutral",
    "diff.high.inverted.angryVSirregular",
    "diff.high.inverted.neutralVSirregular",
    "diff.low.inverted.angryVSneutral",
    "diff.low.inverted.angryVSirregular",
    "diff.low.inverted.neutralVSirregular"
  )
) %>%
  droplevels()

par(mfrow = c(4, 3))

for (i in 1:length(levels(pilot2.posterior.test.regularity$condition))) {
  temp.data <- filter(pilot2.posterior.test.regularity, condition == levels(condition)[i])

  plotPost(
    pull(temp.data, RTs),
    credMass = 0.95,
    compVal = 0,
    ROPE = NULL,
    xlab = "",
    col = "#b3cde0",
    showCurve = FALSE,
    cex = 1,
    main = levels(temp.data$condition)[i]
  )
}

### variability
pilot2.posterior.test.variability <- filter(
  pilot2.posterior.test,
  condition == c(
    "diff.upright.angry.highVSlow",
    "diff.upright.neutral.highVSlow",
    "diff.upright.irregular.highVSlow",
    "diff.inverted.angry.highVSlow",
    "diff.inverted.neutral.highVSlow",
    "diff.inverted.irregular.highVSlow"
  )
) %>%
  droplevels()

par(mfrow = c(2, 3))

for (i in 1:length(levels(pilot2.posterior.test.variability$condition))) {
  temp.data <- filter(pilot2.posterior.test.variability, condition == levels(condition)[i])

  plotPost(
    pull(temp.data, RTs),
    credMass = 0.95,
    compVal = 0,
    ROPE = NULL,
    xlab = "",
    col = "#b3cde0",
    showCurve = FALSE,
    cex = 1,
    main = levels(temp.data$condition)[i]
  )
}

Take-home message:

• paired comparisons:
  • orientation
    • slower RTs for inverted compared to upright irregular faces when within-identity emotion variability is low
  • regularity
    • slightly slower RTs for angry compared to neutral inverted faces when within-identity emotion variability is low
    • slower RTs for irregular compared to neutral inverted faces when within-identity emotion variability is low
  • within-identity emotion variability
    • upright neutral: slower RTs for high compared to low within-identity emotion variability
    • upright irregular: slightly slower RTs for high compared to low within-identity emotion variability

---

---

Exp 1

• flickering frequency: 6 Hz
• regularity: 2 Hz
• 2 face identities
• task: detect dots on faces

# subset data
behav.exp1 <- behav.data %>%
  filter(experiment == "exp1")

exp1.model <- brm(RTs ~ 0 + condition
+ (0 + condition | participant),
data = behav.exp1,
family = gaussian(),
prior = weak.priors,
sample_prior = TRUE,
inits = "random",
control = list(
  adapt_delta = .99,
  max_treedepth = 15
),
chains = num.chains,
iter = num.iter,
warmup = num.warmup,
thin = num.thin,
algorithm = "sampling",
cores = num.chains,
seed = seed.smorfia
)

saveRDS(exp1.model, file = "exp1_model.rds", compress = "gzip")

# subset data
behav.exp1 <- behav.data %>%
  filter(experiment == "exp1")

# load model
exp1.model <- readRDS("exp1_model.rds")

# summary posterior distributions & diagnostics
summary.exp1.model <-
  describe_posterior(exp1.model,
    centrality = "mean",
    ci = 0.95,
    ci_method = "hdi",
    test = NULL,
    diagnostic = "all"
  ) %>%
  mutate(
    Parameter = gsub("b_condition", "", Parameter),
    Parameter2 = gsub("_", " ", Parameter),
    condition = factor(Parameter2,
      levels =
        c(
          "upright angry high", "upright neutral high", "upright irregular high",
          "upright angry low", "upright neutral low", "upright irregular low",
          "inverted angry high", "inverted neutral high", "inverted irregular high",
          "inverted angry low", "inverted neutral low", "inverted irregular low"
        )
    )
  ) %>%
  dplyr::select(Parameter, Mean, HDI95_Low = CI_low, HDI95_High = CI_high, Estimated_Sample_Size = ESS, MonteCarlo_StdErr = MCSE)

kable(summary.exp1.model, digits = 2)

	Parameter	Mean	HDI95_Low	HDI95_High	Estimated_Sample_Size	MonteCarlo_StdErr
10	upright_angry_low	518.3	478.8	558.4	12347	0.18
16	upright_neutral_low	512.2	471.4	553.0	12836	0.18
13	upright_irregular_low	483.6	443.5	523.3	12527	0.18
14	upright_irregular_low	483.6	443.5	523.3	12527	0.18
2	inverted_angry_low	488.6	448.5	528.7	12168	0.19
8	inverted_neutral_low	551.5	510.1	590.3	12425	0.18
5	inverted_irregular_low	538.8	495.8	580.8	12858	0.19
6	inverted_irregular_low	538.8	495.8	580.8	12858	0.19
9	upright_angry_high	504.5	462.3	544.6	13040	0.18
15	upright_neutral_high	511.2	473.9	551.4	12864	0.17
11	upright_irregular_high	562.3	519.7	602.4	12501	0.19
12	upright_irregular_high	562.3	519.7	602.4	12501	0.19
1	inverted_angry_high	525.4	485.1	567.9	12845	0.19
7	inverted_neutral_high	514.6	476.5	555.8	12908	0.18
3	inverted_irregular_high	551.6	508.6	591.5	12503	0.19
4	inverted_irregular_high	551.6	508.6	591.5	12503	0.19

# MCMC chains
exp1.model %>%
  spread_draws(`b_.*`, regex = TRUE) %>%
  dplyr::select(
    "Chain" = ".chain",
    "upright angry high" = "b_conditionupright_angry_high",
    "upright neutral high" = "b_conditionupright_neutral_high",
    "upright irregular high" = "b_conditionupright_irregular_high",
    "upright angry low" = "b_conditionupright_angry_low",
    "upright neutral low" = "b_conditionupright_neutral_low",
    "upright irregular low" = "b_conditionupright_irregular_low",
    "inverted angry high" = "b_conditioninverted_angry_high",
    "inverted neutral high" = "b_conditioninverted_neutral_high",
    "inverted irregular high" = "b_conditioninverted_irregular_high",
    "inverted angry low" = "b_conditioninverted_angry_low",
    "inverted neutral low" = "b_conditioninverted_neutral_low",
    "inverted irregular low" = "b_conditioninverted_irregular_low"
  ) %>%
  as.data.frame() %>%
  mcmc_trace(facet_args = list(ncol = 3)) +
  geom_vline(
    xintercept = seq(0, 2000, 500),
    linetype = "dotted",
    colour = "#999999",
    size = .8,
    alpha = .5
  ) +
  ggtitle("chain convergence") +
  theme_EmoSSR +
  theme(legend.position = "none")

# posterior predictive checks
exp1.model %>%
  posterior_predict(draws = 500) %>%
  ppc_stat_grouped(
    y = pull(behav.exp1, RTs),
    group = pull(behav.exp1, condition),
    stat = "mean"
  ) +
  ggtitle("posterior predictive checks") +
  theme_EmoSSR +
  theme(legend.position = "none")

exp1.posterior.test <- exp1.model %>%
  spread_draws(`b_.*`, regex = TRUE) %>%
  dplyr::select(
    "upright angry high" = "b_conditionupright_angry_high",
    "upright neutral high" = "b_conditionupright_neutral_high",
    "upright irregular high" = "b_conditionupright_irregular_high",
    "upright angry low" = "b_conditionupright_angry_low",
    "upright neutral low" = "b_conditionupright_neutral_low",
    "upright irregular low" = "b_conditionupright_irregular_low",
    "inverted angry high" = "b_conditioninverted_angry_high",
    "inverted neutral high" = "b_conditioninverted_neutral_high",
    "inverted irregular high" = "b_conditioninverted_irregular_high",
    "inverted angry low" = "b_conditioninverted_angry_low",
    "inverted neutral low" = "b_conditioninverted_neutral_low",
    "inverted irregular low" = "b_conditioninverted_irregular_low"
  ) %>%
  mutate(
    ### orientation
    # angry high
    diff.high.angry.uprightVSinverted = `upright angry high` - `inverted angry high`,
    # neutral high
    diff.high.neutral.uprightVSinverted = `upright neutral high` - `inverted neutral high`,
    # irregular high
    diff.high.irregular.uprightVSinverted = `upright irregular high` - `inverted irregular high`,
    # angry low
    diff.low.angry.uprightVSinverted = `upright angry low` - `inverted angry low`,
    # neutral high
    diff.low.neutral.uprightVSinverted = `upright neutral low` - `inverted neutral low`,
    # irregular high
    diff.low.irregular.uprightVSinverted = `upright irregular low` - `inverted irregular low`,
    ### regularity
    ## upright high
    # angry minus neutral
    diff.high.upright.angryVSneutral = `upright angry high` - `upright neutral high`,
    # angry minus irregular
    diff.high.upright.angryVSirregular = `upright angry high` - `upright irregular high`,
    # neutral minus irregular
    diff.high.upright.neutralVSirregular = `upright neutral high` - `upright irregular high`,
    ## upright low
    # angry minus neutral
    diff.low.upright.angryVSneutral = `upright angry low` - `upright neutral low`,
    # angry minus irregular
    diff.low.upright.angryVSirregular = `upright angry low` - `upright irregular low`,
    # neutral minus irregular
    diff.low.upright.neutralVSirregular = `upright neutral low` - `upright irregular low`,
    ## inverted high
    # angry minus neutral
    diff.high.inverted.angryVSneutral = `inverted angry high` - `inverted neutral high`,
    # angry minus irregular
    diff.high.inverted.angryVSirregular = `inverted angry high` - `inverted irregular high`,
    # neutral minus irregular
    diff.high.inverted.neutralVSirregular = `inverted neutral high` - `inverted irregular high`,
    ## inverted low
    # angry minus neutral
    diff.low.inverted.angryVSneutral = `inverted angry low` - `inverted neutral low`,
    # angry minus irregular
    diff.low.inverted.angryVSirregular = `inverted angry low` - `inverted irregular low`,
    # neutral minus irregular
    diff.low.inverted.neutralVSirregular = `inverted neutral low` - `inverted irregular low`,
    ### variability
    ## upright angry
    diff.upright.angry.highVSlow = `upright angry high` - `upright angry low`,
    ## upright neutral
    diff.upright.neutral.highVSlow = `upright neutral high` - `upright neutral low`,
    ## upright irregular
    diff.upright.irregular.highVSlow = `upright irregular high` - `upright irregular low`,
    ## inverted angry
    diff.inverted.angry.highVSlow = `inverted angry high` - `inverted angry low`,
    ## inverted neutral
    diff.inverted.neutral.highVSlow = `inverted neutral high` - `inverted neutral low`,
    ## inverted irregular
    diff.inverted.irregular.highVSlow = `inverted irregular high` - `inverted irregular low`
  ) %>%
  gather(key = "condition", value = "RTs") %>%
  filter(condition == c(
    "diff.high.angry.uprightVSinverted",
    "diff.high.neutral.uprightVSinverted",
    "diff.high.irregular.uprightVSinverted",
    "diff.low.angry.uprightVSinverted",
    "diff.low.neutral.uprightVSinverted",
    "diff.low.irregular.uprightVSinverted",
    "diff.high.upright.angryVSneutral",
    "diff.high.upright.angryVSirregular",
    "diff.high.upright.neutralVSirregular",
    "diff.low.upright.angryVSneutral",
    "diff.low.upright.angryVSirregular",
    "diff.low.upright.neutralVSirregular",
    "diff.high.inverted.angryVSneutral",
    "diff.high.inverted.angryVSirregular",
    "diff.high.inverted.neutralVSirregular",
    "diff.low.inverted.angryVSneutral",
    "diff.low.inverted.angryVSirregular",
    "diff.low.inverted.neutralVSirregular",
    "diff.upright.angry.highVSlow",
    "diff.upright.neutral.highVSlow",
    "diff.upright.irregular.highVSlow",
    "diff.inverted.angry.highVSlow",
    "diff.inverted.neutral.highVSlow",
    "diff.inverted.irregular.highVSlow"
  )) %>%
  droplevels() %>%
  mutate(condition = factor(condition,
    levels = c(
      "diff.high.angry.uprightVSinverted",
      "diff.high.neutral.uprightVSinverted",
      "diff.high.irregular.uprightVSinverted",
      "diff.low.angry.uprightVSinverted",
      "diff.low.neutral.uprightVSinverted",
      "diff.low.irregular.uprightVSinverted",
      "diff.high.upright.angryVSneutral",
      "diff.high.upright.angryVSirregular",
      "diff.high.upright.neutralVSirregular",
      "diff.low.upright.angryVSneutral",
      "diff.low.upright.angryVSirregular",
      "diff.low.upright.neutralVSirregular",
      "diff.high.inverted.angryVSneutral",
      "diff.high.inverted.angryVSirregular",
      "diff.high.inverted.neutralVSirregular",
      "diff.low.inverted.angryVSneutral",
      "diff.low.inverted.angryVSirregular",
      "diff.low.inverted.neutralVSirregular",
      "diff.upright.angry.highVSlow",
      "diff.upright.neutral.highVSlow",
      "diff.upright.irregular.highVSlow",
      "diff.inverted.angry.highVSlow",
      "diff.inverted.neutral.highVSlow",
      "diff.inverted.irregular.highVSlow"
    )
  ))

# summary
summary.exp1.posterior.test <- exp1.posterior.test %>%
  group_by(condition) %>%
  dplyr::summarize(
    mean = mean(RTs),
    hdi.95.low = hdi(RTs)[1],
    hdi.95.high = hdi(RTs)[2],
    evidence.ratio = ifelse(mean < 0, 
                            length(which(RTs < 0)) / length(which(RTs > 0)),
                            length(which(RTs > 0)) / length(which(RTs < 0))
    )
  ) %>%
  ungroup()

kable(summary.exp1.posterior.test, digits = 2)

condition	mean	hdi.95.low	hdi.95.high	evidence.ratio
diff.high.angry.uprightVSinverted	-21.36	-80.45	40.29	3.30
diff.high.neutral.uprightVSinverted	-2.83	-61.94	54.26	1.11
diff.high.irregular.uprightVSinverted	11.26	-44.92	67.51	1.90
diff.low.angry.uprightVSinverted	28.80	-23.00	86.58	4.85
diff.low.neutral.uprightVSinverted	-39.60	-94.69	18.74	10.31
diff.low.irregular.uprightVSinverted	-55.50	-111.17	7.23	25.64
diff.high.upright.angryVSneutral	-5.05	-57.45	46.57	1.38
diff.high.upright.angryVSirregular	-58.29	-114.48	-7.75	65.70
diff.high.upright.neutralVSirregular	-50.36	-104.63	4.11	30.76
diff.low.upright.angryVSneutral	5.51	-54.08	57.62	1.41
diff.low.upright.angryVSirregular	36.40	-13.24	95.86	9.25
diff.low.upright.neutralVSirregular	27.69	-33.34	76.91	4.65
diff.high.inverted.angryVSneutral	9.66	-52.89	64.12	1.76
diff.high.inverted.angryVSirregular	-26.21	-92.34	28.23	3.69
diff.high.inverted.neutralVSirregular	-37.44	-93.08	19.43	8.67
diff.low.inverted.angryVSneutral	-62.82	-119.87	-7.30	82.38
diff.low.inverted.angryVSirregular	-50.13	-109.73	2.73	22.82
diff.low.inverted.neutralVSirregular	10.48	-51.57	65.02	1.78
diff.upright.angry.highVSlow	-14.61	-78.51	38.20	2.13
diff.upright.neutral.highVSlow	-1.05	-56.25	59.14	0.96
diff.upright.irregular.highVSlow	79.98	23.94	140.32	332.50
diff.inverted.angry.highVSlow	38.25	-25.66	89.45	11.81
diff.inverted.neutral.highVSlow	-36.27	-89.68	15.85	9.93
diff.inverted.irregular.highVSlow	11.19	-42.50	70.65	1.85

### Paired comparisons ###

### orientation
exp1.posterior.test.orientation <- filter(
  exp1.posterior.test,
  condition == c(
    "diff.high.angry.uprightVSinverted",
    "diff.high.neutral.uprightVSinverted",
    "diff.high.irregular.uprightVSinverted",
    "diff.low.angry.uprightVSinverted",
    "diff.low.neutral.uprightVSinverted",
    "diff.low.irregular.uprightVSinverted"
  )
) %>%
  droplevels()

par(mfrow = c(2, 3))

for (i in 1:length(levels(exp1.posterior.test.orientation$condition))) {
  temp.data <- filter(exp1.posterior.test.orientation, condition == levels(condition)[i])

  plotPost(
    pull(temp.data, RTs),
    credMass = 0.95,
    compVal = 0,
    ROPE = NULL,
    xlab = "",
    col = "#b3cde0",
    showCurve = FALSE,
    cex = 1,
    main = levels(temp.data$condition)[i]
  )
}

### regularity
exp1.posterior.test.regularity <- filter(
  exp1.posterior.test,
  condition == c(
    "diff.high.upright.angryVSneutral",
    "diff.high.upright.angryVSirregular",
    "diff.high.upright.neutralVSirregular",
    "diff.low.upright.angryVSneutral",
    "diff.low.upright.angryVSirregular",
    "diff.low.upright.neutralVSirregular",
    "diff.high.inverted.angryVSneutral",
    "diff.high.inverted.angryVSirregular",
    "diff.high.inverted.neutralVSirregular",
    "diff.low.inverted.angryVSneutral",
    "diff.low.inverted.angryVSirregular",
    "diff.low.inverted.neutralVSirregular"
  )
) %>%
  droplevels()

par(mfrow = c(4, 3))

for (i in 1:length(levels(exp1.posterior.test.regularity$condition))) {
  temp.data <- filter(exp1.posterior.test.regularity, condition == levels(condition)[i])

  plotPost(
    pull(temp.data, RTs),
    credMass = 0.95,
    compVal = 0,
    ROPE = NULL,
    xlab = "",
    col = "#b3cde0",
    showCurve = FALSE,
    cex = 1,
    main = levels(temp.data$condition)[i]
  )
}

### variability
exp1.posterior.test.variability <- filter(
  exp1.posterior.test,
  condition == c(
    "diff.upright.angry.highVSlow",
    "diff.upright.neutral.highVSlow",
    "diff.upright.irregular.highVSlow",
    "diff.inverted.angry.highVSlow",
    "diff.inverted.neutral.highVSlow",
    "diff.inverted.irregular.highVSlow"
  )
) %>%
  droplevels()

par(mfrow = c(2, 3))

for (i in 1:length(levels(exp1.posterior.test.variability$condition))) {
  temp.data <- filter(exp1.posterior.test.variability, condition == levels(condition)[i])

  plotPost(
    pull(temp.data, RTs),
    credMass = 0.95,
    compVal = 0,
    ROPE = NULL,
    xlab = "",
    col = "#b3cde0",
    showCurve = FALSE,
    cex = 1,
    main = levels(temp.data$condition)[i]
  )
}

Take-home message:

• paired comparisons:
  • orientation
    • slightly slower RTs for inverted compared to upright neutral faces when within-identity emotion variability is low
    • slightly slower RTs for inverted compared to upright irregular faces when within-identity emotion variability is low
  • regularity
    • slower RTs for irregular compared to angry upright faces when within-identity emotion variability is high
    • slightly slower RTs for irregular compared to neutral upright faces when within-identity emotion variability is high
    • slower RTs for neutral compared to angry inverted faces when within-identity emotion variability is low
    • slightly slower RTs for irregular compared to angry inverted faces when within-identity emotion variability is low
  • within-identity emotion variability
    • upright irregular: slower RTs for high compared to low within-identity emotion variability
    • inverted angry: slightly slower RTs for high compared to low within-identity emotion variability

---

---